Half-Time SPECT Myocardial Perfusion Imaging with Attenuation ...

Half-Time SPECT Myocardial Perfusion
Imaging with Attenuation Correction
Iftikhar Ali 1 , Terrence D. Ruddy 1,2 , Abdulaziz Almgrahi 1 , Frank G. Anstett 3 , and R. Glenn Wells 1,2
1 Division of Cardiology, University of Ottawa Heart Institute, Ottawa, Ontario, Canada; 2 Division of Nuclear Medicine, The Ottawa
Hospitals, University of Ottawa, Ottawa, Ontario, Canada; and 3 GE Healthcare, Waukesha, Wisconsin
Reducing acquisition time may improve patient throughput, increase
camera efficiency, and reduce costs; reducing acquisition
time also increases image noise. Newly available software controls
the effects of noise by maximum a posteriori reconstruction while
maintaining resolution with resolution-recovery methods. This
study compares half-time (HT) gated myocardial SPECT images
processed with ordered-subset expectation maximization with
resolution recovery (OSEM-RR) (with and without CT-based attenuation
correction [AC]) with full-time (FT) images obtained with a
standard clinical protocol and reconstructed with filtered backprojection
(FBP) and OSEM (with and without AC). Methods: Atotalof
212 patients (mean age, 57 y; age range, 27–86 y) underwent 1-d
rest/stress 99mTc-tetrofosmin gated SPECT. FT (12.5 min, both
rest and stress) and HT (rest, 7.5 min; stress, 6.0 min) images
were acquired with low-dose CT for AC in 112 patients. HT acquisitions
were processed with OSEM-RR (with and without AC) using
software, and FT acquisitions were processed with FBP and
OSEM (with and without AC). In another 100 patients, test–retest
repeatability was assessed using 2 sets of FT images (FBP reconstruction)
that were acquired one immediately after the other. Radiologists
unaware of the acquisition and reconstruction protocols
visually assessed all reconstructed images for summed stress,
summed rest, and summed difference scores and regional wall
motion using a 17-segment model. Automated analysis on gated
SPECT was used to determine left ventricular volumes, ejection
fraction, and dilation (end-diastolic volume, end-systolic volume,
left ventricular ejection fraction, and transient ischemic dilation
[TID]). A clinical diagnosis was also determined. Results: All measurements
resulted in significant correlations (P , 0.01) between
the HT and FT images. The only significant difference in mean
values was for OSEM-RR plus AC; this method led to an increase
in TID by 4% over FT imaging. The concordance in the clinical diagnosis
for HT versus FT was 106 to 112 (k 5 0.88) for no AC and
102 to 106 (k 5 0.91) for AC, similar to the repeatability of FT versus
FT (98/100, k 5 0.95). Conclusion: HT images processed with the
new algorithm provided a clinical diagnosis in concordance with
that from FT images in 95% (no AC) to 96% (AC) of cases. This concordance
is similar to the test–retest repeatability of FT imaging.
Key Words: SPECT; myocardial perfusion imaging; resolution
recovery; attenuation correction
J Nucl Med 2009; 50:554–562
DOI: 10.2967/jnumed.108.058362
Received Sep. 22, 2008; revision accepted Dec. 8, 2008.
For correspondence or reprints contact: R. Glenn Wells, Division of
Cardiology Cardiac Imaging, First Floor, University of Ottawa Heart
Institute, 40 Ruskin St., Ottawa, ON, K1Y 4W7, Canada.
E-mail: gwells@ottawaheart.ca
COPYRIGHT ª 2009 by the Society of Nuclear Medicine, Inc.
554 THE JOURNAL OF NUCLEAR MEDICINE • Vol. 50 • No. 4 • April 2009
Coronary heart disease is a leading cause of death in the
United States and the most prevalent form of cardiovascular
disease, affecting 16 million people. Gated SPECT cardiac
perfusion imaging plays a fundamental role in prediction,
diagnosis, prognosis, and the selection of appropriate treatment
for patients with suspected or known ischemic heart
disease (1,2).
In SPECT, there is a trade-off between image-acquisition
time and noise levels. A shorter acquisition time would
reduce the time during which the patient is required to lie still
and reduce the likelihood of patient motion artifacts. A
shorter scan time might also improve camera use and reduce
the cost of the test. There was some success in reducing scan
times with the introduction of multihead g-cameras in the
1990s, which increased the detector surface area and allowed
simultaneous acquisition of multiple-projection views. Other
groups have suggested that diagnostic accuracy could be
maintained with noisier studies (3) or studies with reduced
spatial resolution (4), but these approaches had limited
acceptance. More recently, there have been exciting advances
in image processing and reconstruction algorithms
that improve the image quality of count-poor studies.
The new algorithms incorporate noise regularization and
resolution recovery (RR). To compensate for the increased
noise in the shorter-duration acquisitions, some form of
statistically based noise suppression, based on prior knowledge
of the type of activity distribution, is applied to the data
during image reconstruction. This form of noise suppression,
like low-pass filtering, tends to reduce the spatial resolution
of the image. To compensate for these effects, the algorithms
also include distance-dependent RR. The resolution of gcamera
images decreases as the source increases its distance
from the collimator surface. By modeling the point-spread
function (PSF) and other sources of image blurring inside the
image-creation process, it is possible to compensate for these
effects and improve the image quality (5). The PSF can be
integrated into reconstruction as part of an iterative algorithm.
The PSF has 4 components: septal penetration, septal
scatter, intrinsic detector response, and, most important,
geometric response function (GRF) (6). The GRF component
is defined as the normalized 2-dimensional Fourier transform
of the point-source image formed in a plane behind the

collimator by photons that are accepted geometrically
through the collimator holes. It is often assumed that GRF
is spatially invariant in planes parallel to the face of the
collimator and that the intrinsic response is spatially invariant
in the detection plane. However, a more accurate description
of specific collimators can be obtained by Monte Carlo
simulation or experimental measurements of collimator
geometry (7,8).
Our objective in this study was to compare gated myocardial
perfusion images obtained using half-time (HT)
acquisition and processed using one of the new RR reconstruction
algorithms with those obtained using a standard
clinical protocol and reconstructed using filtered backprojection
(FBP) or ordered-subset expectation maximization
(OSEM) in a patient population of intermediate-to-high
pretest probability of coronary artery diseases (CAD).
MATERIALS AND METHODS
HT Reconstruction Algorithm
The Evolution for Cardiac software package (GE Healthcare) is
a new reconstruction algorithm that has recently become available
for clinical cardiac imaging. This algorithm incorporates RR and
maximum a posteriori (MAP) noise regularization, permitting
cardiac SPECT images to be acquired in approximately half of the
time (HT acquisition) needed for reconstruction with the standard
OSEM algorithm (9,10). The PSF is stored as a look-up table, and
the radial distance of the detector is recorded as part of the
projection data. Equipment-related information for a variety of
different collimators is stored in an additional look-up table that is
part of the reconstruction package; acquisition parameters are
TABLE 1. Study Population Demographics
retrieved from all projection data. The RR process can produce
noise hot spots that detrimentally affect image quality. The MAP
algorithm is used to suppress the formation of hot spots. It is a
1-step late implementation using a Green’s prior with parameters
that have been tuned for each clinical protocol and for gated and
attenuation-corrected images (11). Attenuation correction (AC)
with Evolution for Cardiac also includes scatter correction based
on a dual-energy window. A median root prior is used in the last
iteration of the reconstruction to preserve edge sharpness (12).
This scheme has been found to give an image of equal or superior
quality on limited clinical studies and on physical phantom data
(13–20).
Patient Population
We prospectively evaluated 212 patients from July 2007
through March 2008 at the Nuclear Cardiology Diagnostic Imaging
laboratory, University of Ottawa Heart Institute, Ottawa,
Canada. Patients were selected from those who were scheduled
to have routine rest/stress myocardial perfusion SPECT for
suspected or known CAD. A total of 112 patients underwent
full-time (FT) and HT acquisitions. Another 100 patients underwent
2 FT acquisitions to evaluate test–retest reproducibility. The
patients had a mean age of 57 y (age range, 27–86 y), and their
clinical characteristics are provided in Table 1. The test–retest
group had more males and a higher pretest likelihood of CAD.
However, most cases in both groups were of moderate risk
(10%290%). The study was approved by the University of
Ottawa’s Human Ethics Review Board, and patients gave written
consent before enrollment.
Radiopharmaceutical Injection and Image Acquisition
99m Tc-tetrofosmin (Myoview; GE Healthcare) was used in all
the studies. The radiopharmaceutical doses were administered,
Patient characteristic FT–HT (n 5 112) FT–FT (n 5 100) P
Male 67 (60%) 74 (74%) 0.03
Age (y) NS
Mean 58 56
Range 32–86 27–80
Body mass index (mean 6 SD) 28.3 6 4.4 28.3 6 4.0 NS
Hypertension 74 (66%) 58 (58%) NS
Dyslipidemia 62 (55%) 56 (56%) NS
Smoking (current and previous) 59 (53%) 54 (54%) NS
Diabetes 18 (16%) 14 (14%) NS
Family history of premature CAD 46 (41%) 41 (41%) NS
Previous event
Myocardial infarction 23 (21%) 20 (20%) NS
Percutaneous coronary intervention 19 (17%) 25 (25%) NS
Coronary artery bypass graft 10 (9%) 8 (8%) NS
Pretest likelihood of CAD (combined
Diamond/Forrester)
Low risk (,10%) 42 (38%) 20 (20%) 0.004
Moderate risk (10%290%) 63 (56%) 73 (73%) 0.009
High risk (.90%) 7 (6%) 7 (7%) NS
Stress test
Exercise 53 (47%) 56 (56%) NS
Dipyridamole 59 (53%) 44 (44%) NS
Data in parentheses are percentages unless otherwise noted.
NS 5 not significant.
HALF-TIME SPECT MPI WITH AC • Ali et al. 555

and images were acquired according to the guidelines of the
American Society of Nuclear Cardiology (ASNC) (21). A 1-d rest
and stress imaging protocol was used in all patients. For rest
images, the injected dose of 99m Tc-tetrofosmin was, on average,
344 MBq (9.3 mCi), with a range of 286–445 MBq (7.7–12 mCi).
For the stress images, the average injected dose was 1,053 MBq
(28.5 mCi), with a range of 930–1,281 MBq (25.1–34.6 mCi). All
rest images were acquired at 45 min after tetrofosmin injection,
and all stress images were acquired between 30 and 60 min after
tetrofosmin injection.
Exercise and Dipyridamole Stressing Protocols
Patients were advised, at the discretion of their referring physician,
to stop taking nitrates, calcium antagonists, and b-blockers at
the time of the myocardial perfusion study. They were also asked
to refrain from ingesting caffeine-containing foods and beverages,
drugs, and dipyridamole for 24 h before their study.
Unless contraindicated or at the request of the referring physician,
all patients underwent a symptom-limited Bruce protocol
treadmill exercise test. Targeted heart rate was 85% of maximum
age-adjusted predicted heart rate (22). At peak stress, the radiotracer
was injected intravenously, and exercise was continued for
an additional 60 s. The patients who were unable to achieve the
targeted heart rate or exercise were stressed pharmacologically.
Dipyridamole (0.142 mg/kg/min) was infused intravenously for
5 min, and 99m Tc-tetrofosmin was injected at 2 min after infusion
completion (7 min into the study). Aminophyline (100–200 mg)
was given intravenously 2 min after injection of the radiotracer.
Data Acquisition and Reconstruction
All studies were completed with a 90°-angled dual-head integrated
SPECT/CT camera (Infinia Hawkeye 4; GE Healthcare)
using standard (FT) and short (HT) protocols. All images were
acquired with a low-energy high-resolution collimator using a
15% energy window centered on the 140-keV photopeak of 99m Tc,
based on ASNC guidelines for cardiac 99m Tc studies (21). In the
standard protocol, SPECT data were acquired for 25 s per
projection (30 projections/head; total of 180° of data) for both
the rest and the stress studies. In the short protocol, data were
acquired for 15 s per projection for the rest study and 10 s per
projection for the stress study. Both rest and stress studies were
gated at 8 frames per cardiac cycle. In one half of the 112 patients,
the HT scan was acquired first, followed by the FT scan; in the
other half of the patients, the scan order was reversed. A single
low-dose CT scan (helical, 140 keV, 1 mA with 1.9 pitch) was
acquired at the end of both rest and stress acquisitions for AC.
Vendor-supplied quality-assurance software was used to ensure
the CT scan was registered with both HT and FT SPECT images.
Projections and reconstructed matrix size for all scans were 64 ·
64 pixels, with a pixel size of 6.8 · 6.8 mm. Reconstruction was
done on a Xeleris 2.0 (GE Healthcare) workstation running on
xw6200 hardware (Hewlett Packard). For the 100 patients undergoing
the test–retest portion of the study, 2 standard SPECT scans
were obtained, one immediately after the other and with the same
acquisition parameters as described above.
All FT scans were reconstructed with both FBP and OSEM.
Both gated and ungated images were used for evaluation. The FBP
images were created with a 10th-order Butterworth filter, using
different cutoff frequencies for the ungated (0.4 cycles/cm) and
gated (0.35 cycles/cm) data. The OSEM iterative reconstruction (2
iterations, 10 subsets) was done both with CT-based AC (IRAC)
556 THE JOURNAL OF NUCLEAR MEDICINE • Vol. 50 • No. 4 • April 2009
and without CT-based AC (IRNC). The OSEM images were also
filtered after reconstruction with a 10th-order Butterworth filter:
ungated (0.4 cycles/cm) and gated (0.35 cycles/cm).
All HT scans were reconstructed with Evolution for Cardiac
using manufacturer-recommended RR and noise-reduction parameters
(12 iterations, 10 subsets) both with CT-based AC (IRACRR)
and without CT-based AC (IRNCRR). The factory-recommended
Butterworth filters were applied to the ungated rest (0.4 cycles/
cm), ungated stress (0.52 cycles/cm), and gated rest and stress (0.4
cycles/cm) data.
The resulting images were randomized and analyzed in a
masked manner by 2 experienced American Board of Nuclear
Medicine–certified nuclear medicine physicians using 4DM-
SPECT (Invia Medical Imaging Solutions) on a Hermes Gold
workstation (Hermes Medical Solutions). Differences between the
readers were resolved with consensus.
Data Collection and Statistical Analysis
The reconstructed images were assessed subjectively for image
quality as poor, good, or excellent. The summed stress score (SSS)
(graded from 0 [normal tracer uptake] to 4 [absent tracer]),
summed rest score (SRS), and regional wall motion (graded as
normal, hypokinetic, akinetic, or dyskinetic) were visually assessed
using a 17-segment model. Automated analysis on gated
SPECT was used to determine the end-diastolic volume (EDV),
end-systolic volume (ESV), left ventricular ejection fraction
(LVEF), and transient ischemic dilation (TID). Finally, a clinical
diagnosis was determined from perfusion, TID, and wall motion
as normal, probably normal, equivocal, probably abnormal, or
abnormal. Fixed defects with normal wall motion were considered
to be normal perfusion with an attenuation artifact. For this study,
an SSS greater than or equal to 2 and a TID greater than 1.2 (23)
were considered abnormal. A more typical cutoff value for normal
SSS is less than 4 (24), but we have chosen a stricter criterion of
less than 2 for this study to emphasize any minor differences in
apparent perfusion.
The concordance of clinical diagnosis for the FT and HT
images was evaluated using the intraclass correlation coefficient
(ICC-r) and k-coefficient. The SSS, SRS, summed difference
scores (SDS), and quantitative parameters (LVEF, LVEDV, and
LVESV) were compared using linear regression analysis, the ICC-r,
and Bland–Altman analysis. Mean values were compared using a
Bonferroni-corrected 2-tailed paired t test, with a value of P less
than 0.05 considered significant.
RESULTS
FBP Analysis
There were no differences in subjective image quality
among the FT images reconstructed with FBP and IRNC.
Image quality was excellent in 111 of 112 images reconstructed
with both FBP and IRNC, and the last image was
adequate on IRNC. No significant difference in SSS, SRS,
SDS, and wall motion were observed between FBP and IRNC
(Supplemental Table 1; supplemental materials are available
online only at http://jnm.snmjournals.org). Quantitative
analysis of left ventricular volumes (LVEDV, LVESV, and
LVEF) also showed no significant difference between FBP
and IRNC. With respect to clinical diagnosis, there were 2 of
112 (1.8%) discordant cases. Both cases were normal on FBP.

The patient in the first discordant case had mild ischemia in
the right coronary artery territory on IRNC, and subsequent
82 Rb-PET findings were normal. The patient in the second
discordant case also had mild ischemia in the left circumflex
artery territory on IRNC but was unavailable for follow-up.
This level of disagreement is similar to the discordance seen
in the test–retest repeatability.
For the 100-patient test–retest portion of the study, all
measures (SSS, SRS, SDS, wall motion, LVEDV, LVESV,
LVEF, and TID) showed significant correlation and no
significant differences in the mean values (Supplemental
Table 1). Bland–Altman analysis of the LV volumes and
ejection fraction showed no changes in the distribution of
differences as a function of mean values. The clinical
diagnosis of the 100 patients from the test–retest portion
of this study showed excellent reproducibility: ICCr 5 0.98
(P 5 0.001). The patient in the first discordant case had
mild-to-moderate ischemia in the inferior wall in the initial
acquisition. However, in the subsequent acquisition there
was scatter from the gut that led to a diagnosis of subdiaphragmatic
attenuation, with normal wall motion. The
patient in the second discordant case had normal perfusion
in both acquisitions but had a change from normal-tomoderate
global hypokinesis.
Visual Analysis
No differences in subjective image quality between FT
and HT images were observed (Fig. 1). Image quality was
excellent in 110 of 112 for HT images reconstructed with
IRNCRR, 112 of 112 for HT images reconstructed with
IRACRR, and 111 of 112 for FT images reconstructed
with both IRNC and IRAC. Highly significant correlations
for SSS and SRS were observed between IRNCRR and
IRNC (ICCr 5 0.94 for both SSS and SRS) and between
IRAC and IRACRR (ICCr 5 0.96 for both) (Supplemental
Table 2). The correlation for the SDS measure was reduced
FIGURE 1. Stress/rest 99m Tc-tetrofosmin SPECT study. FT acquisition reconstructed with OSEM without AC (IRNC) (A), HT
acquisition reconstructed with OSEM and RR without AC (IRNCRR) (B), FT acquisition reconstructed with OSEM with CT-based
AC (IRAC) (C), and HT acquisition reconstructed with OSEM and RR with CT-based AC (IRACRR) (D). No difference in image
quality or clinical diagnosis was observed in 95% of cases between IRNC and IRNCRR and 96% of cases between IRAC and
IRACRR. S 5 stress; R 5 rest.
HALF-TIME SPECT MPI WITH AC • Ali et al. 557

FIGURE 2. Histogram analysis comparing
HT and FT myocardial perfusion
SPECT reconstructed with (IRAC vs.
IRACRR) and without (IRNC vs.
IRNCRR) CT-based AC. Results for
test–retest repeatability (FBP1 vs.
FBP2) are also shown. SSS (A) and
SDS (B) are presented separately.
(without AC, ICCr 5 0.55; with AC, ICCr 5 0.39). The
reduced correlation was caused by the large number of
patients with an SDS of 0 and the small range of SDS
differences in our patient sample (range, 25 to 4). The
concordance in SDS remained good; the SDS differences
between FT and HT images (Fig. 2) were 0 in 95 of 112
patients without AC and 96 of 106 patients with AC and
were greater than or equal to 4 in only 3 of 112 patients
without AC and 2 of 106 patients with AC.
Quantitative Analysis of LV Volume and Function
Quantitative analyses were performed on both FT and
HT gated stress acquisitions. We compared HT and FT
images without (IRNCRR and IRNC) and with CT-based
AC (IRACRR and IRAC) and found no significant differences
in LVEDV, LVESV, and LVEF (Supplemental Table
2). LVEF determined from HT SPECT correlated well with
that of FT without AC (IRNC vs. IRNCRR, ICCr 5 0.90)
and with CT-based AC (IRACRR vs. IRAC, ICCr 5 0.88)
(Fig. 3). The Bland–Altman plot analysis of IRNCRR with
IRNC and IRACRR with IRAC showed no significant
differences in LVEF (Fig. 4). Similar results were seen
for LVEDV and LVESV (Supplemental Table 2). The only
difference observed was a small but significant (P 5 0.04)
increase in the average TID for IRACRR (TID 5 0.96),
compared with IRAC (TID 5 0.92).
Clinical Diagnosis
Clinical diagnoses obtained from IRNCRR were 95%
(106/112) concordant with IRNC, and the k-coefficient was
0.88. Clinical diagnoses obtained from IRACRR were 96%
(102/106) concordant with IRAC, and the k-coefficient was
0.91. The differences in SSS in the 6 discordant cases were
mild; they were within the range of 2 to 4. Frequently an
SSS greater than or equal to 4 is required for a scan to be
considered abnormal. In this study, we use an SSS cutoff of
2. We also used a TID value greater than or equal to 1.2 as
abnormal and considered fixed defects with normal wall
motion to be attenuation artifacts (for a scan without AC).
The discordant clinical diagnoses (Table 2) are detailed in
Table 3. CT-based AC resolved all of the differences in
perfusion (cases 2–6) (Fig. 5). In case 1, the patient was
claustrophobic and a CT scan could not be acquired. An
558 THE JOURNAL OF NUCLEAR MEDICINE • Vol. 50 • No. 4 • April 2009
82 Rb-PET follow-up study showed normal perfusion. Patients
in cases 5 and 6 were also followed up with an
electrocardiogram-gated 82 Rb-PET scans, and both PET
studies had normal perfusion and normal TID. The patient
in case 7 underwent a subsequent angiogram during the
course of clinical evaluation, which showed no significant
stenosis supplying the inferolateral wall, the area of ischemia
noted in both no-AC studies and the IRAC study. The
patient in case 8 relocated and was unavailable for further
follow-up.
DISCUSSION
This study compared images obtained with HT acquisition
and processed with the Evolution for Cardiac reconstruction
algorithm with conventional image acquisition
and processing protocols using FBP and OSEM. Our results
indicate that the new reconstruction algorithm with RR can
reduce acquisition time and provide similar image quality
and diagnostic accuracy for myocardial perfusion in terms
of perfusion, wall motion, and volumes for the evaluation
FIGURE 3. Linear regression analysis of LVEF comparing
HT and FT myocardial perfusion SPECT with (IRAC vs.
IRACRR) and without (IRNC vs. IRNCRR) CT-based AC.

of CAD. Moreover, these results are similar both with and
without the application of CT-based AC. Mild discrepancies
in FT–HT are similar to test–retest repeatability of
FT imaging.
The depth-dependant RR approach has been adapted by
various vendors in their new reconstruction algorithms with
various methods of modeling components: Evolution for
Cardiac (10,13,15,20), Wide-Beam Reconstruction (WBR)
(UltraSPECT) (10,13,15,16), Astonish (Philips Medical System)
(17,18), and 3D Flash (Siemens Medical Solutions) (19).
Our study supports and expands on the results and conclusions
of other work done previously with the Evolution for Cardiac
package. DePuey et al. compared 2 new reconstruction algorithms
with FBP. These authors demonstrated that for cardiac
perfusion SPECT, HT Evolution for Cardiac (OSEM-RR) and
WBR offered stress and rest image quality and diagnostic
accuracy equivalent to FT FBP (13,15). DePuey et al. also
demonstrated that these new methods (OSEM-RR and WBR)
offered gated myocardial perfusion SPECT functional image
quality with HTacquisitions equivalent to or possibly superior
to that achieved with FT acquisition processed with FBP (10).
In the studies by these authors, images were acquired on a
Ventri 90°-angled dual-head scintillation camera (GE Healthcare),
and we have demonstrated similar results here on an
Infinia Hawkeye (GE Healthcare). DePuey et al. compared HT
OSEM-RR and WBR with FT FBP using metrics of image
quality, LV volumes, LVEF, and wall motion. In our study, we
extended these results to clinical diagnosis, SSS, SRS, SDS,
and CT-based AC. We had also compared 2 sets of FT
acquisitions reconstructed with a clinically used FBP algorithm
for test–retest reproducibility.
This study suggests that CT-based AC may improve
concordance between HT and FT images. In those cases for
which there was discordance between the IRNC and
IRNCRR results, the inclusion of CT-based AC removed
the differences in perfusion. A possible explanation is that
the addition of AC in the algorithm improves the accuracy of
the camera over RR alone and leads to greater consistency in
the data and a more accurate reconstructed image. Narayanan
et al. demonstrated that iterative reconstruction provides
better performance than does FBP reconstruction for the
overall detection of CAD and the localization of perfusion
defects in the 3 arterial territories (25). They noted a steady
improvement in defect detection as attenuation, scatter, and
FIGURE 4. Bland–Altman analysis of
LVEF comparing HT and FT myocardial
perfusion SPECT without (IRNC vs.
IRNCRR) (A) and with (IRAC vs.
IRACRR) (B) CT-based AC. LVEF (IRNC
vs. IRNCRR): mean difference, 0.1; 1.96
SDs, 9.4. LVEF (IRAC vs. IRACRR):
mean difference, 0.4; 1.96 SDs, 11.0.
resolution losses were progressively compensated for in the
reconstruction, suggesting that improved modeling of physical
degradations can improve the quality of cardiac perfusion
images. Thus, use of both resolution and attenuation
compensation might be expected to yield a better result than
would resolution compensation alone, although with the
small number of cases in this study the differences were not
shown to be significant.
A 100-patient test–retest study was performed on the same
camera with the same methodology to determine the reproducibility
for specific parameters (SSS, SRS, SDS, LVEDV,
LVESV, LVEF, and wall motion) and clinical diagnosis
through masked reading of separate studies performed one
immediately after the other. This study yielded a discrepancy
rate for the clinical diagnosis of 2%. Although this rate is
slightly lower than the discrepancy rate seen between the FT
and HT images, the small test–retest variability would
account for some of the discrepancy seen, making the true
difference between the HT and FT images even less.
In this study, we observed a small but significant (P 5 0.04)
increase of 4% in the average TID for IRACRR, compared
with IRAC. TID may be caused by an apparent thinning of the
myocardial wall due to reduced uptake in the endocardium
(26). TID is measured on the basis of the change in LV volume
between rest and stress images. Resolution recovery has the
potential to alter the spatial resolution of images and thereby
influence the accuracy of edge detection and measured vol-
TABLE 2. Clinical Diagnosis Without and With
CT-Based AC
HT
FT
No AC
Normal Abnormal
Normal 69 5
Abnormal
AC
1 37
Normal 71 2
Abnormal 2 31
Normal studies had normal perfusion, TID, and wall motion.
Abnormal studies had SSS greater than or equal to 2 or TID
greater than 1.2 or wall motion abnormalities greater than or
equal to 2.
HALF-TIME SPECT MPI WITH AC • Ali et al. 559

TABLE 3. Discordance in Clinical Diagnosis Between HT and FT Images
No AC (IRNC vs. IRNCRR) CT-based AC (IRAC vs. IRACRR)
Case no. Perfusion (SDS) TID WM Perfusion TID WM
1 X (2, 0) *
2 X (0, 3)
3 X (0, 4)
4 X (0, 2)
5 X (0, 2) X (1.03, 1.25)
6 X (0, 1) X (1.03, 1.3) X (1.04, 1.27)
7 X
8 X
*CT was unavailable (because of patient’s claustrophobia).
Differences in perfusion were mild in all cases (SDS of #4). In cases of discordant TID results, values of TID for 2 image sets are noted
in parentheses.
umes. In light of this and the observed increase in TID, it may
be necessary to redefine the reference range for TID using the
HT reconstruction algorithm.
A disadvantage of the new reconstruction algorithm is
that there is a significant increase in image reconstruction
time with RR (11 min for each IRNCRR and IRACRR),
FIGURE 5. Stress/rest 99m Tc-tetrofosmin SPECT study. FT acquisition reconstructed with OSEM without AC (IRNC) shows
breast attenuation in anterior wall (A), HT acquisition reconstructed with OSEM and RR without AC (IRNCRR) shows reversible
defect in anterior wall (B), FT acquisition reconstructed with OSEM with AC (IRAC) shows normal perfusion (C), and HT
acquisition reconstructed with OSEM and RR with CT-based AC (IRACRR) also shows normal perfusion (D). S 5 stress
tetrofosmin study; R 5 rest tetrofosmin study.
560 THE JOURNAL OF NUCLEAR MEDICINE • Vol. 50 • No. 4 • April 2009

compared with standard clinical reconstruction (1–2 min
for FBP and 2–3 min for IRNC). The increase in processing
time might have the potential to affect workflow, offsetting
the reduced acquisition time. However, for the purposes of
image-quality assurance, the projection data are immediately
available. In addition, the total reconstruction time is
still considerably less than camera time required for each
patient when the time for patient positioning and camera
set-up are included. Thus, reconstructed images are completed
before the next patient’s projection data become
available and no backlog in processing should occur.
Study Limitations
The RR and noise reduction described in this study for HT
acquisition are modeled specifically for each camera and
collimator; thus, the result of 1 camera cannot be completely
generalized to other types of cameras and collimators. The
result of the study described in this report is validated only for
the Infinia Hawkeye 4 SPECT/CT dual-head camera. Therefore,
care should be taken in extending these results to other
systems, and further validation may be necessary.
The filters used for the FT images were different from
those recommended for the HT images. Incorporating RR
and MAP priors into the reconstruction alters the noise and
resolution properties of the resultant images, making it
unlikely that the optimal filters for the new algorithm will
coincide with those currently used as optimal for our clinical
(FT) images. In this study, we chose to use the manufacturerrecommended
postfiltering parameters. These are expected
to be a reasonable choice for filtering but may not represent
the optimal choice for our patient population. The differences
in filtering combined with RR may alter the definition of the
myocardial wall and could, therefore, alter the automatic
computation of the valve plane or ventricular volume, for
example. The filters could possibly be adjusted to improve
further the already high concordance between the FTand HT
images. In addition, some studies have shown that imaging
early after stress (,30 min) can result in transient wall
motion abnormalities (27). This effect may have contributed
to differences seen in some cases because of the serial nature
of our imaging protocol.
Differences were noted between some FT and HT images
for which follow-up was not available. This was true for
images both with and without AC, although there was a trend
toward a smaller number of discrepancies with the inclusion
of AC. Because of the lack of an independent standard, it
remains uncertain which of the approaches yields a more
accurate result in this small number of cases. Additionally,
the total number of discrepancies between the HT and FT
images is small, making it difficult to draw any meaningful
conclusion as to whether the images from one approach are
superior to those of the other. Further investigation with an
independent standard such as PET would be necessary to
settle this question, and the results are unlikely to be clinically
meaningful as the number of studies in question is 4% or
fewer.
Clinical Relevance
The advantages of reducing acquisition time by 50% are
considerable and include improvement in image quality
through reduced patient motion, improvement in patient
experience with less discomfort and anxiety, improvement
in patient throughput, and, finally, possible increase in
camera efficiency. Alternatively, the radiopharmaceutical
dose could be halved while maintaining a FT acquisition, to
decrease radiation exposure to patients and staff.
CONCLUSION
A new OSEM-based reconstruction algorithm incorporating
depth-dependant RR and MAP noise suppression has
been evaluated in 112 patients both with and without CTbased
AC. HT images processed with the new algorithm
provided a clinical diagnosis in concordance with the diagnosis
from FT images processed with standard algorithms in
95% (no AC) to 96% (with AC) of cases. The concordance of
the FT–HT imaging is similar to test–retest repeatability of
FT imaging.
ACKNOWLEDGMENTS
We gratefully acknowledge the technical assistance of
Patti M.E. Irvine, Rene Adeny, Elaine Cooper, Brian F.
Marvin, Marlie A. Poirier, and Lyanne N. Fuller. We thank
GE Healthcare for providing the Evolution for Cardiac
software package. This study was funded in part by the
Ontario Research Fund (ORF)-Research Excellence Program.
One author is supported by a research fellowship
funded by the University of Ottawa, Ottawa, Canada.
REFERENCES
1. DePuey EG, Rozanski A. Using gated technetium-99m-sestamibi SPECT to
characterize fixed myocardial defects as infarct or artifact. J Nucl Med. 1995;
36:952–955.
2. Thomas GS, Miyamoto MI, Morello AP, et al. Technetium 99m sestamibi
myocardial perfusion imaging predicts clinical outcome in the community
outpatient setting: the Nuclear Utility in the Community (NUC) Study. J Am Coll
Cardiol. 2004;43:213–223.
3. Taillefer R, Primeau M, Lambert R, Levelle Y. Technetium-99m-sestamibi
myocardial perfusion imaging: comparison between a short (8 minutes) and
standard (21 minutes) data acquisition time in diagnosis of coronary artery
disease [abstract]. J Nucl Med. 1992;33(suppl):855P.
4. DePuey EG, Nichols KJ, Slowikowski JS, et al. Fast stress and rest acquisitions
for technetium-99m sestamibi separate day SPECT. J Nucl Med. 1995;36:569–
574.
5. Gifford HC, King MA, Wells RG, Hawkins WG, Narayanan MV, Pretorius PH.
LROC analysis of detector response compensation in SPECT. IEEE Trans Med
Imaging. 2000;19:463–473.
6. Frey EC, Tsui BMW. Collimator-detector response compensation in SPECT.
In: Zaidi H, ed. Quantitative Analysis in Nuclear Medicine Imaging. Berlin,
Germany: Springer; 2005.
7. Metz CE, Atkins FB, Beck RN. The geometric transfer function component for
scintillation camera collimators with straight parallel holes. Phys Med Biol.
1980;25:1059–1070.
8. Tsui BMW, Gullberg GT. The geometric transfer-function for cone and fan beam
collimators. Phys Med Biol. 1990;35:81–93.
9. Tsui BMW, Frey EC, Zhao X, Lalush DS, Johnston RE, McCartney WH. The
importance and implementation of accurate 3D compensation methods for
quantitative SPECT. Phys Med Biol. 1994;39:509–530.
HALF-TIME SPECT MPI WITH AC • Ali et al. 561

10. DePuey G, Gadiraju R, Clark J, Thompson L, Anstett F, Schwartz SC. Ordered
subset expectation maximization and wide beam reconstruction ‘‘half-time’’
gated myocardial perfusion SPECT functional imaging: a comparison to ‘‘fulltime’’
filtered backprojection. J Nucl Cardiol. 2008;15:547–563.
11. Green PJ. Bayesian reconstruction from emission tomography data using a
modified EM algorithm. IEEE Trans Med Imaging. 1990;9:84–93.
12. Alenius S, Ruotsalainen U. Bayesian image reconstruction for emission
tomography based on median root prior. Eur J Nucl Med. 1997;24:258–265.
13. DePuey G, Gadiraju R, Anstett F, et al. Half-time gated myocardial perfusion
SPECT processed with both OSEM with resolution recovery and wide beam
reconstruction: clinical performance compared to full-time filtered backprojection.
In: Proceedings of the ICNC meeting; May 2, 2007; Prague, Czech
Republic. Abstract 472.
14. Belhocine T, Gambhir S, Brenner R, Peretz A, Driedger A, Urbain JL. Half-time
resolution recovery package for SPECT-CT MPI: a pilot study [abstract]. J Nucl
Med. 2007;48(suppl):234P.
15. DePuey G, Gadiraju R, Anstett F. OSEM and WBR half-time gated myocardial
perfusion SPECT: a comparison to filtered backprojection [abstract]. J Nucl
Med. 2007;48(suppl):237P.
16. Borges-Neto S, Pagnanelli RA, Shaw LK, et al. Clinical results of a novel wide
beam reconstruction method for shortening scan time of Tc-99m cardiac SPECT
perfusion studies. J Nucl Cardiol. 2007;14:555–565.
17. Heller GV, McGhie AI, Philips RD, et al. Multi-center evaluation of a new
reconstruction method with depth dependent collimator resolution and scatter
compensation applied to half-time small field of view SPECT MPI. Paper
presented at: ASNC 2007 meeting; September 6–9; San Diego, California.
18. Bateman TM, Heller GV, McGhie AI. Application of simultaneous Gd-153 line source
attenuation correction to half-time SPECT acquisitions: multi-center clinical evaluation.
Paper presented at: ASNC 2007 meeting; September 6–9; San Diego, California.
562 THE JOURNAL OF NUCLEAR MEDICINE • Vol. 50 • No. 4 • April 2009
19. Kritzman JN, Cahill JM, Ficaro EP, Corbett JR. Quantitative comparison of
standard and reduced acquisition time attenuation corrected and non-corrected
myocardial perfusion images: a phantom study utilizing a 3D iterative reconstruction
[abstract]. J Nucl Cardiol. 2007;14(suppl 1):S107–S108.
20. Valenta I, Treyer V, Husmann L, et al. New reconstruction algorithm allows
shortened acquisition time for myocardial perfusion SPECT [abstract]. J Nucl
Med. 2008;49(suppl 1):72P.
21. Hansen CL, Goldstein RA, Akinboboye OO, et al. Myocardial perfusion and
function: single photon emission computed tomography. J Nucl Cardiol. 2007;
14:e39–e60.
22. Henzlova MJ, Cerqueira MD, Mahmarian JJ, Yao SS. Stress protocols and
tracers. J Nucl Cardiol. 2006;13:e80–e90.
23. Abidov A, Bax JJ, Hayes SW, et al. Transient ischemic dilation ratio of the left
ventricle is a significant predictor of future cardiac events in patients with
otherwise normal myocardial perfusion SPECT. J Am Col Cardiol. 2003;42:
1818–1825.
24. Iskandrian AE. Risk assessment of stable patients (panel III) [abstract]. J Nucl
Cardiol. 1999;6:93.
25. Narayanan MV, King MA, Pretorius PH, et al. Human-observer receiveroperating-characteristic
evaluation of attenuation, scatter, and resolution compensation
strategies for 99m Tc myocardial perfusion imaging. J Nucl Med.
2003;44:1725–1734.
26. Iskandrian AS, Heo J, Nguyen T, Lyons E, Paugh E. Left ventricular dilatation
and pulmonary thallium uptake after single-photon emission computer tomography
using thallium-201 during adenosine-induced coronary hyperemia. Am J
Cardiol. 1990;66:807–811.
27. Giorgetti A, Rossi M, Stanislao M, et al. Feasibility and diagnostic accuracy of a
gated SPECT early imaging protocol: a multicenter study of the Myoview
imaging optimization group. J Nucl Med. 2007;48:1670–1675.